Getting started

Connect LoreKit to your AI coding agent in three steps — read and write memories so knowledge accumulates across sessions.

Generate an API token

After signing in, go to Settings → API keys and click Generate new token. Give it a name you'll recognise (e.g. claude-work, cursor-home) and choose a permission tier:

lk_rw_… — read + write (recommended for coding agents)
lk_ro_… — read only (context injection, CI)
lk_wo_… — write only (one-way memory feeders)

The token is shown once — copy it before closing the modal.

Add LoreKit to your agent's MCP config

Replace lk_rw_… with your token and drop the snippet into your agent's config file.

.mcp.json

{
  "mcpServers": {
    "lorekit": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://pqokxlhvnosogizsjztg.supabase.co/functions/v1/mcp", "--header", "Authorization:Bearer lk_rw_…"]
    }
  }
}

Project-local. Add .mcp.json to .gitignore — it contains your token.

Tip:All five agents use the same URL format — only the config file path differs. Requires Node.js; npx downloads mcp-remote automatically on first run.

Verify the connection

Start a new session in your agent and ask it to write a test memory:

memory.write {
  scope: "global",
  key:   "hello-lorekit",
  value: "Connection is working."
}

If the write succeeds, the Lore Explorer (in the dashboard) will show the memory. You can also run the CLI health check:

npx @lorekit/cli doctor

Optional: auto-memories from PR reviews

Connect a GitHub webhook and every resolved PR review comment becomes a memory automatically, tagged source::pr-webhook. Set this up any time from Settings → Integrations after signing in.

Note:Next steps: see the Remote storage tutorial for token tiers and CI injection, or Team sharing to share lore across your whole team.